Vincent Gracco, School of Communication Sciences and Disorders and Centre for Research on Brain, Language & Music, McGill University, Montreal, Canada, and Haskins Laboratories, New Haven, CT

A comprehensive understanding of the functional and dysfunctional neural mechanisms for speech production is critical to theory and clinical practice. Current neural models of speech production rely to a significant extent on empirical data obtained from functional magnetic resonance imaging (fMRI) studies. However, for the most part, systems-level neural interactions for speech instantiated in current models and theory are incompletely specified due to an almost exclusive focus on a single component of the neurovascular signal, the positive blood-oxygenation-level-dependent (BOLD) response (PBR). The PBR is associated with excitatory neurotransmitters, increased local field potentials and task-positive networks (TPN). In contrast, a ubiquitous component of all neurovascular activity that has received less attention is the negative BOLD response or NBR. Based on data from both humans and nonhumans the NBR and the associated task negative network (the Default Mode Network or DMN) reflects a suppressive or inhibitory process. The DMN, however, is a functionally heterogeneous network made up of core nodes that can be functionally decoupled, the extent to which is beginning to be examined in detail. Importantly, the appropriate balance between positive and negative network activity is a hallmark of normal neural function impacting developmental processes as well as maintaining a homeostatic balance in developed networks. Here we provide some empirical data to highlight aspects of task-positive and task-negative BOLD interactions during speech production. We use a combination of spatial Independent Components Analysis (sICA) and General Linear Modeling (GLM) to extract functional networks and assess their overlap and contribution to tasks ranging from simple repetition of words and sentences to more cognitive tasks involving response selection. The results are used to highlight the importance of including contributions of task deactivation and the NBR in neural models of speech production.